Official PyTorch implementation of the paper:
DFAN++: Enhanced Triple-branch Network for Generalized Zero-Shot Image Classification
Neurocomputing 2025
DFAN++ Framework: Integrating EAFL, CDM and CPP branches
| Dataset | CZSL (acc) | GZSL (U) | GZSL (S) | GZSL (H) |
|---|---|---|---|---|
| CUB | 78.7 | 72.9 | 76.8 | 74.8 |
| SUN | 68.1 | 51.4 | 40.5 | 45.3 |
| AWA2 | 71.7 | 66.2 | 84.5 | 74.2 |
# Install dependencies
pip install -r requirements.txt | Dataset | Download Link | Attributes | Classes | Images |
|---|---|---|---|---|
| CUB | download link | 312 | 200 | 11788 |
| SUN | download link | 102 | 717 | 14340 |
| AWA2 | download link | 85 | 50 | 37322 |
| aPY | download link | 64 | 32 | 15399 |
we use standard splits from Xian et al.. We provide trained models on three different datasets: (https://pan.baidu.com/s/1cfN3lu-_fTOc8LaOhmv1Uw?pwd=1111) Extraction code：1111
@article{ZHU2025130880,
title = {DFAN++: Enhanced triple-branch network for generalized zero-shot image classification},
journal = {Neurocomputing},
volume = {653},
pages = {130880},
year = {2025},
issn = {0925-2312},
doi = {https://doi.org/10.1016/j.neucom.2025.130880},
url = {https://www.sciencedirect.com/science/article/pii/S0925231225015528},
author = {Yifan Zhu and Lei Xiang and Yuan Zhou and Haoran Duan and Yang Long} Parts of our codes based on:
If you have any questions about codes, please don't hesitate to contact us by yifanzhu12138@gmail.com.